Divide and conquer: Separating the two probabilities in seismic phase picking

There are two fundamental probabilities in the seismic phase picking process – the probability of the existence of a seismic phase (detection probability) and the probability of correctly identifying the phase arrival time (timing probability). The nearly ubiquitous approach in developing deep learning phase picking models is to use a kernel, such as a truncated Gaussian, to mask the labeled phase arrival time, and train a segmentation model. Once a model is trained, the times of the peaks in the output are taken as phase arrival times (picks) and the height of the peaks are taken as “probability” of the picks. Here, we show that this “probability” represents neither the detection nor the timing probabilty because this approach forces the output to follow the shape of the kernel. We introduce an approach using two models to estimate these two distinct probabilities. We use a binary classifier with a calibrated confidence to address the detection probability and a multi-class classifier to obtain a probability mass function to address the timing probability. This new approach makes the deep learning-based phase picking process more interpretable and gives us options to logically control seismic monitoring workflows.